[openstack-qa] Ganularity of 'gate' tag

David Kranz david.kranz at qrclab.com
Thu May 16 12:28:44 UTC 2013


There are valid arguments both ways and I don't have a strong opinion 
about this. But we should state a policy and be
consistent about it, or at least provide some guidance to new test writers.

  -David

On 5/16/2013 7:17 AM, Sean Dague wrote:
> On 05/15/2013 03:45 PM, David Kranz wrote:
>> I had thought this tag would mostly be at the class level but I see
>> patches that are instead putting it on every method of a class.
>> Is there a good reason to do that? Given that we are trying to put as
>> much as possible in the gate I'm not sure the case of mixed 
>> gate/non-gate
>> methods in the same class will be common.
>
> That's possible, though previously all methodology was function level 
> tagging not class level. It has it's draw backs, but it also adds the 
> advantage that new tests wouldn't be gate by default.
>
> I can go either way. My goal is we can run a big set, which isn't 
> everything. However we get there I'm ok with, but the fact that we can 
> only easily run smoke (which is small enough to not find a lot of 
> issues), and everything in the tree, I see as a problem.
>
> But I won't be able to really sort through it all in my own head until 
> the restructure is done, as there is a volume race that's been exposed 
> just by changing test order. 
> http://logs.openstack.org/28624/15/check/gate-tempest-devstack-vm-quantum/22583/logs/screen-c-sch.txt.gz 
> - 2013-05-15 15:00:05
>
> That's preventing moving out the scenario tests.
>
>     -Sean
>




More information about the openstack-qa mailing list